search for: trytovector

On Aug 20, 2013, at 14:49 , Nadav Rotem <nrotem at apple.com> wrote: > Hi Matt, > > This code maintains a vector of float4 and it inserts and extracts values from this vector. The ’select’ operations are already vectorized. Maybe a sequence of inst-combines (or DAG-combines) can help. If you re-write this code using scalars then the slp-vectorizer, with some tweaks, will be able

...d29 = add nsw i32 %add27, %15 store i32 %add29, i32* %sum, align 4, !tbaa !1 ret void } As observed above, SLP vectorization doesn't vectorize this code. I tried debugging, just mentioning what i found - the code hits vectorizeChainsInBlock() -> if(ShouldStartVectorizeHorAtStore) -> tryToVectorize() -> tryToVectorizePair() -> tryToVectorizeList() - inside this function, it iterates in the for loop for checking if scalar instructions are of same type. It exits for 2nd iteration in the loop as OpCode does not match , one opcode is 'add' and second is 'load'. So it exit...

...\ / > > \ / > > + > > | > > | > > ↓ > > > > sum > > > > > > This doesn't vectorize, because in function tryToVectorizeList(), we > check if the > > > > scalar instruction are of the same type. In above case, it will be fadd > and load, > > > > and hence no vectorization takes place. > > > > > > > > case 2: > > > > float a[4], sum; > > void f...

Hi Nadav, Thanks for the quick reply !! Ok, so as of now we are lacking capability to handle flat large reductions. I did go through function vectorizeChainsInBlock() (line number 2862). In this function, we try to vectorize if we have phi nodes in the IR (several if's check for phi nodes) i.e we try to construct tree that starts at chains. Any pointers on how to join multiple trees? I